Running a program in gdb, but ignoring a graceful exit?

Question

We're running squid from with gdb - that way we can automatically generate backtraces for debugging.

backtrace=`mktemp`
gdb -q -x /etc/service/squid3/gdbcommands /usr/sbin/squid 2>&1 >$backtrace
/usr/bin/mail -s "`hostname`: Squid was restarted (backtrace)" [email protected] < backtracetrace
rm $backtrace

/etc/service/squid3/gdbcommands contains:

 set args -NsYC
 handle SIGPIPE pass nostop noprint
 handle SIGTERM pass nostop noprint
 handle SIGUSR1 pass nostop noprint
 handle SIGHUP  pass nostop noprint
 handle SIGSEGV stop
 handle SIGABRT stop
 run
 set print pretty
 backtrace full
 generate-core-file
 quit

But, every now and then, squid is "just" being stopped & restarted, with no crash being involved at all. In that case I'm still getting an email containing:

  Reading symbols from /usr/sbin/squid...done.
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
  [Inferior 1 (process 57867) exited normally]
  /etc/service/squid3/gdbcommands:10: Error in sourced command file:
  No stack.
  (gdb) quit

And of course there's no stack, since the program exited ok.

How can I change my gdbcommands file to avoid this?

Answer 1

This can be done using either Python or the gdb CLI. Since the CLI is a bit simpler, when possible, I'll sketch that approach.

First, you might as well only create a core file on a bad exit. And, we'll use the gdb exit code later, so let's arrange for that to tell the calling script what happened.

Where your current script says:

backtrace full
generate-core-file
quit

... instead use:

if !$_isvoid($_exitsignal) || (!$_isvoid($_exitcode) && $_exitcode != 0))
  backtrace full
  generate-core-file
  quit 0
end
quit 1

Then your calling script can check the exit code of gdb:

if gdb your args here; then
   mail results
fi